PSI - Issue 28

Marouene Zouaoui et al. / Procedia Structural Integrity 28 (2020) 978–985 Marouene Zouaoui et al. / Structural Integrity Procedia 00 (2019) 000–000

982

5

As previously announced, this work is part of developing a full numerical model of the pre-structured material that goes until fracture. Thus reaching the yielding point with the right stress and the right strain levels is crucial. Since an elastic non-linearity was observed, calculating Young’s modulus as the slope of the tangent to the true stress-strain curve at low stress will lead to an under estimations of the strain at the yielding point. This is why another method for calculating Young’s modulus was used. It consists of considering a linear elastic behavior until the yielding point calculated with an offset of 0.2 % strain. 4. Finite element model validation In order to check the validity of the identified elastic constants, an oriented numerical model for each specimen will be created. The local references in mesh element assignment method described in [4] will be used to reproduce the printing orientation inside the model. The numerical answer will be compared to collect data using the DIC analysis. The cost function was formulated based on the Finite Element Model Updated widely used [13]–[15] that represents an inverse method to identify the inputs of a numerical model. The validity of those inputs is related to their capacity the reproduce experimental results (fields measurement) with a low level of error. The objective here is to evaluate the difference between the numerical and experimental forces as well as the difference between the DIC and numerical strain fields while using the identified material’s constants ( , , , � ,except that was proved to have no influence on the formulated cost function. 4.1. Subsets and Integration points linking A linking procedure was performed between the subsets of the reference snapshot and the integration points of the undeformed mesh. The density of subsets is higher than the integration points, so it was decided to reduce the experimental data size to fit the number of integration points by linking each one to the neighbor subsets within a radius less than half the used mesh seize. The following figure shows the reference snapshot localized on the undeformed finite element mesh of the tensile specimen.

Fig. 3. Reference snapshot linked with the undeformed mesh

4.2. Loading steps choice Several loading steps were chosen from the start of the test until the yielding point. Such choice was mad to ensure the prediction accuracy of the mechanical behavior of the pre-structured material. After computing the cost function for each loading step it was decided to discard the steps at low levels of strain and load. Such choice was made knowing that low strain values are easily perturbed by the noise and the measured load values are affected by the loading phase (slip between the jaws).